May 19, 2010 –
Netflix, the popular movie-delivery site, “predicts” films its users may enjoy. The process doesn’t involve mindreading, however; it’s simply computer science.
A collaborative filtering system allows Netflix to match users’ viewing and rating histories to others with similar profiles, resulting in the ability to hone in on which movies might appeal to which customers.
Informatics professor and CALIT2 academic affiliate Bill Tomlinson and his research group use collaborative filtering, too, although in a slightly more altruistic way. Informatics doctoral student Joel Ross and ICS undergraduate Nitin Shantharam incorporated the technique into a Web-based carbon footprint calculator, allowing it to more accurately “predict” users’ impacts on the environment based on the answers to a few quick questions.
Most carbon-footprint calculators provide estimates based on answers to a lengthy list of sometimes cumbersome questions. Users must respond to detailed queries about their homes, utilities expenditures, airline travel mileage, cars, food choices and recycling efforts.
The Social Code Group’s Better Carbon calculator (www.bettercarbon.com), by contrast, estimates carbon footprints based primarily on the location of the user’s home, information that yields a lot more data about a person’s lifestyle than one might expect. “There’s an illusion in America that we’re all unique but much of the data we’re gathering is not that hard to predict,” Tomlinson states.
A home address narrows down many factors that contribute to the user’s carbon footprint: amount of air travel, likely driving habits and even dietary choices. “We are able to make an educated guess based simply on a user’s location,” he says.
While the initial estimate is admittedly rough, it can be refined by plugging in answers to some of the same questions found on other carbon calculators. The Better Carbon advantage, however, is that users don’t have to answer all the questions to get an accurate estimate; each answer refines the final calculation – bit by bit – by more closely matching the user to others with similar responses. The collaborative filtering techniques provide carbon footprint estimates that are comparable to those with similar lifestyles.
Better Carbon was designed to be faster and easier than its brethren so that more people will actually use it to determine their environmental impact.
And the more users the site attracts, the more accurate it becomes. Researchers launched the site with data collected from 400 people, but each time someone enters information, that data further filters the resulting estimates.
“The more information the system gets, the finer and finer the [distinctions] become,” Tomlinson says.
Flexibility is a key component. As environmental research progresses, researchers can add new questions to the calculator. Because all questions on other sites must be answered before a prediction can be made, those calculators must limit the number of queries they present.
Better Carbon, on the other hand, can eventually address 60 or 70 factors, or more. “With our system, you can answer just one question, or as many as you want. If even a couple of people answer a question, it will influence the rest of the system and help us produce estimates for those values,” Ross says. “So for the price of 10 questions we can give you answers to 50.”
The Better Carbon site includes a chart that details what percentage of a user’s footprint is attributed to specific lifestyle choices; it also incorporates a social action component. Users are asked to specify one action they take to reduce their carbon footprint and those recommendations are sorted so others can see suggestions from those whose lives are similar to their own. “The site can show you where your problem areas are, and others’ recommendations can give you ideas as to how to address similar problems,” Tomlinson says. Better Carbon is also available through a Facebook app, allowing users to compare their carbon footprints with those of their friends.
“A carbon calculator is an educational and informational tool that can allow people to see ways to adjust their own lives or to start taking broader political action,” Tomlinson says. Adds Ross: “You can find out which of your behaviors have the largest impact, and then act on that information. We’re trying to improve that process.”
The researchers’ paper on the project was accepted to this week’s 2010 IEEE International Symposium on Sustainable Systems and Technology in Washington D.C.
— Anna Lynn Spitzer